Nonlinear Models for Determining Mode Choice
Due to the increasing complexity in transportation systems, one needs to search for different ways to model the separate components of these systems. A general transportation system comprises components/models concerning mode choice, travel duration, trip
- PDF / 170,294 Bytes
- 12 Pages / 430 x 660 pts Page_size
- 17 Downloads / 209 Views
2
Transportation Research Institute, Hasselt University, Science Park 5/6, 3590 Diepenbeek, Belgium [email protected], [email protected] Center for Statistics, Hasselt University, University Campus - Building D, 3590 Diepenbeek, Belgium [email protected]
Abstract. Due to the increasing complexity in transportation systems, one needs to search for different ways to model the separate components of these systems. A general transportation system comprises components/models concerning mode choice, travel duration, trip distance, departure time, accompanying individuals, etc. This paper tries to discover whether semi- and nonlinear models bring an added value to transportation analysis in general and mode choice modelling in particular. Linear (logistic regression), semi-linear (multiple fractional polynomials) and nonlinear (support vector machines and classification and regression trees) models are applied to several binary settings and compared to each other based on sensitivity (i.e. the proportion of positive cases that are predicted correctly). In general, one can state that on skewed data sets, linear and semi-linear models tend to perform better, whereas on more balanced data sets both nonlinear models yield better results. Future research will take a closer look at other extensions of the well-established linear regression model.
1
Introduction
At the beginning of the 21st century, transportation planners are facing an increasing complexity in trying to predict travel demand. On the one hand, transportation systems are inherently very complex systems, including a large number of components that needs to be dealt with, such as efficient, safe and reliable transportation, but also the impact on the environment and on the surrounding communities. Above all this, the economic growth gives rise to an ever-increasing travel demand, creating even more problems (Sadek, 2007). On the other hand, at the same time, different algorithms and models have been developed in different research fields, such as artificial intelligence, statistics, machine learning, etc. The transport analysts are thus faced with a wide variety of types of models to choose from to model the separate components of the transportation systems. It is within this context that artificial intelligence (AI) paradigms can and need to be used to address some of the aforementioned problems that are quite challenging to solve by means of traditional and classical solution methods. This paper J. Neves, M. Santos, and J. Machado (Eds.): EPIA 2007, LNAI 4874, pp. 183–194, 2007. c Springer-Verlag Berlin Heidelberg 2007
184
E. Moons, G. Wets, and M. Aerts
focusses on one of the most important application areas of AI within transportation, i.e. decision making, which happens multiple times every day when a trip is planned. With whom do I undertake this trip, when do I carry it out, where do I go to, what will the duration be, which transport mode do I take, etc.? In this paper, the focus lies on the latter decision, i.e. the mode choice. However, t
Data Loading...
